The present invention is in the technical field of semiconductors. In particular, this invention relates to highly integrable edge emitting active optical devices such as lasers, optical amplifiers and the like, and the manufacture of the same.
For over forty years, silicon has been the material of choice for high density microelectronics in large part because of the performance advantages of high speed, low static power complementary metal oxide semiconductor (CMOS) technology. With the maturity of silicon fabrication processes gained over this time, and the ever-increasing prominence of silicon devices in the marketplace, a significant area of research in the field of optoelectronics has been to integrate both active and passive optoelectronic devices directly on silicon substrates.
Silicon based compounds are regularly used to make passive optical devices such as waveguides, splitters, couplers, and wave division multiplexers. In fact, silicon, and silicon based materials, such as silicon dioxide, silicon nitride, and silicon oxy-nitride, are widely used materials used in commercial optical planar waveguides today. The development of silicon active optical devices like lasers and optical amplifiers, however, has proved to be much more challenging. The difficulty lies in the fact that silicon is an inefficient light emitting material due to its indirect energy band gap. Silicon laser research efforts have investigated ways to circumvent this limitation by using materials such as nanoporous silicon (See, e.g., Kojima et al., Applied Physics Letters 87 (2005)), rare-earth doped silica glasses (See, e.g., deWaal et al., Applied Physics Letters 71, 2922 2924 (1997); deWaal et al., IEEE Phot. Tech. Letters 16, 194 196 (2004)), silicon nano-crystals (See, e.g., Pavesi et al., Nature 408, 440 444 (2000)) and strained germanium on silicon (See, e.g., Michel et al., IEEE J. Select. Topics in Quant. Electron. 12, 1628 1635 (2006)), or have exploited phenomena such as the Raman Effect (See, e.g., Rong et al., Nature 433, 725 727 (2005)). Although these demonstrations represent tremendous breakthroughs, each of these approaches requires an additional laser to pump the devices and achieve light emission. To make silicon the material of choice for monolithic optoelectronic integration, the development of an efficient electrically pumped active optical device is necessary.
Given the difficulty of manufacturing electrically pumped silicon based active optical devices of sufficient quality, an alternative approach involving the hybrid integration of III-V semiconductor and silicon substrates has been considered. One prior art approach has involved the epitaxy, chemical vapor deposition, or growth of III-V based semiconductors directly onto a silicon substrate. This growth process is then commonly followed by processing that attempts to transform the semiconductor material into a working active optical device. A common problem with this approach is that there are both lattice constant and thermal expansion coefficient mismatches between the III-V based semiconductor and silicon that result in significant stresses in the semiconductors. Alternatively, another prior art approach has involved the bonding of III-V based semiconductors directly onto silicon substrates followed by processing that attempts to transform the semiconductor material into a working active optical device. With this approach, the thermal expansion coefficient mismatch is a common problem since the bonding occurs at elevated temperatures.
A key requirement for any process used to manufacture an active optical device, such as an edge emitting laser, is that it results in smooth, end facets. Smooth facets help to minimize problems such as facet heating and promote ideal characteristics such as low current threshold operation. The ideal laser facet is one that has been cleaved along one of the natural cleavage planes of the semiconductor crystal from which the laser is made. The location of cleavage planes in semiconductors can be due to a variety of factors, such as favorable atomic densities along certain planes (e.g., Silicon) or electrical surface neutrality conditions (e.g., Gallium Arsenide, Indium Phosphide) (See, e.g., Siernans et al., Phys. Rev. B, 59(4):3000 3007 (1999)). (100) InP and (100) GaAs substrates, upon which the majority of semiconductor lasers are grown, most easily cleave in two dimensions, one that is parallel to the plane of the wafer flat, and the other dimension which is perpendicular to the wafer flat.
The facets for active optical devices, such as lasers, optical amplifiers, and the like, are conventionally formed using cleaving tools. These cleaving tools use scribing and/or mechanical forces to cause the semiconductor to smoothly break along the cleavage plane. Unfortunately, these techniques can locate the cleave position to a precision of no better than 5 microns (See, e.g., Marsh et al., J. of Crystal Growth, 288, Iss. 1, 2-6 (2006)), which severely hampers control of the laser cavity length. The control of the length of a laser cavity is extremely important in many applications. For instance, the laser cavity length can affect the optical emission properties of the device. In addition, in cases where the goal is to integrate the active optical devices with other devices, it is often desirable to have very precise control of the device dimensions. As an example, one way to integrate laser diodes and other active semiconductor devices within silicon-based photonic integrated circuits (PICs) is to use micro-scale hybrid integration and co-axial alignment. This approach involves fabricating in-plane active optical device building blocks, such as laser diodes and optical amplifiers, in the form of thin platelets and assembling these thin devices in dielectric recesses formed to intersect waveguides in a PIC fabricated on silicon wafers, or other substrates. By coaxially aligning the active optical devices with the waveguides of the PIC, very efficient coupling between the active devices and the photonic circuitry can be achieved. This concept is illustrated in FIG. 27, which shows the integration of an edge emitting laser block 23 with a dielectric waveguide 24 on silicon 25.
Low-loss coaxial coupling involves the direct alignment of the active device and waveguide in both the horizontal and vertical planes with little or no gap between the active device and waveguide facets. The dimensions of the waveguide, active optical device, and recess formed on the silicon platform must all be precisely controlled since the alignment accuracy is the major determinant of the resulting coupling efficiency.
The vertical offset between the active optical device and dielectric waveguides can easily be kept small, because the layers that comprise the dielectric waveguides and the semiconductor active optical devices can all be deposited with precise thickness control by, for example, using plasma enhanced chemical vapor deposition (PECVD) to deposit the dielectrics, and metal organic chemical vapor deposition (MOCVD) or molecular beam epitaxy (MBE) to grow the semiconductor heterostructures. The horizontal offset between the non-light emitting, receiving, or transmitting sides of laser and waveguide is also straightforward to control because the widths of the dielectric recesses and the laser platelets can be controlled precisely using modern photolithography and vertical dry etching. The length of the recesses is similarly straightforward to control.
The more difficult challenge with the coaxial coupling integration strategy is the accurate and consistent dimensioning of the length of the edge emitting laser which allows the gap between the laser and waveguide facets to be kept small. As mentioned earlier, conventional cleaving tools used to produce laser facets can locate the cleave with a precision of no better than ±5 microns, which in turn means the laser cavity length can vary as much as 20 microns. Since the recess must accommodate the longest devices, some conventionally cleaved device platelets will be this much shorter than the recess.
An additional problem with conventional cleaving methods is that the mechanical nature of these cleaving processes does not allow for reliable cleaving of thin-film semiconductors (i.e., tens of microns or thinner). Thin film semiconductor devices are especially desirable for integration with other devices.
An alternative to cleaving facets, is to etch low loss highly reflective optical facets. In fact, several companies, as mentioned in Behfar et al., SPIE Optoelectronics Magazine, 27 29 (2005) have commercialized their facet etching processes. The main drawback with facet etching, though, is that unlike cleaving, etching seems to invariably impart roughness on the facets and does not result in an ideal facet.